03 — Regression Modeling
Simple and multivariable linear regression with residual diagnostics and short-window prediction generalization.
chiller-plant regression scikit-learn prediction
Pythonscikit-learnscipymatplotlib
Key Findings
- ● Single-variable regression (kW vs OAT): R²=0.86, MSE=5,055
- ● Multivariable regression (OAT, tons, occupancy): R²=0.98, MSE=1,099
- ● 24-hour training window generalizes well to full dataset (R²=0.94)
Progressive regression modeling of chiller power consumption: from single-variable OAT regression to multivariable models with residual analysis. Tests whether a short 24-hour training window can generalize across the full operating period.